Abstract
Modern natural language processing models such as transformers operate multimodal data. In the present paper, multimodal data is explored using multimodal topic modeling on transactional data of bank corporate clients. A definition of the importance of modality for the model is proposed on the basis of which improvements are considered for two modeling scenarios: preserving the maximum amount of information by balancing modalities and automatic selection of modality weights to optimize auxiliary criteria based on topic representations of documents.
A model is proposed for adding numerical data to topic models in the form of modalities: each topic is assigned a normal distribution with learning parameters. Significant improvements are demonstrated in comparison with standard topic models on the problem of modeling bank corporate clients. Based on the topic representations of the bank’s customers, a 90-day delay on the loan is predicted.
Notes
A matrix \(F \in \mathbb {R}^{m \times n} \) is said to be stochastic if \(F_{ij} \geqslant 0 \) and \(\sum \nolimits _{i = 1}^m F_{ij} = 1\), so that the columns form probability distributions.
Unimodal representations are obtained using the M-step for a unimodal topic model with one modality with the same value of \(p_{tdw} \).
The cardinality of a modality is the number of tokens of that modality in the document.
REFERENCES
Vaswani, A., Shazeer, N., Parmar, N., et al., Attention is all you need, Adv. Neural Inf. Process. Syst., 2017, vol. 30.
Devlin, J., Chang, M., Lee, K., and Toutanova, K., BERT: Pre-training of deep bidirectional transformers for language understanding, Proc. 2019 Conf. North Am. Ch. Assoc. Comput. Linguist.: Human Lang. Technol. Vol. 1 (Long and Short Pap.) (2019), pp. 4171–4186.
Zhu, W., Tao, D., Cheng, X., et al., BERT4Rec: Sequential recommendation with bidirectional encoder representations from transformer, CIKM, 2019, pp. 1441–1450.
Pavlovski, M., Gligorijevic, J., Stojkovic, I., et al., Time-aware user embeddings as a service, Proc. SIGKDD Conf. Knowl. Discovery Data Min. (2020).
Reynolds, D., Gaussian Mixture Models, Boston: Springer, 2009.
Hoffman, T., Probabilistic latent semantic indexing, Proc. 22nd Annu. Int. ACM SIGIR Conf. Res. Dev. Inf. Retr. (SIGIR’99) (1999), pp. 50–57.
Dempster, A.P., Laird, N.M., and Rubin, D.B., Maximum likelihood from incomplete data via the EM algorithm, J. R. Stat. Soc. Ser. B, 1977, vol. 39, pp. 1–38.
Kuhn, H.W. and Tucker, A.W., Nonlinear programming, Proc. Second Berkeley Symp. Math. Stat. Probab. (1950), pp. 481–492.
Ke, G., Meng, Q., Finley, Th., et al., LightGBM: A highly efficient gradient boosting decision tree, Adv. Neural Inf. Process. Syst., 2017, vol. 30.
Paszke, A., Gross, S., Massa, F., et al., PyTorch: An imperative style, high-performance deep learning library, Adv. Neural Inf. Process. Syst., 2019, vol. 30.
Vorontsov, V. and Potapenko, A., Additive regularization of topic models, Mach. Learn., 2015, vol. 101, pp. 303–323.
Funding
This work was supported by the Russian Foundation for Basic Research, project no. 20-07-00936.
Author information
Authors and Affiliations
Corresponding authors
Additional information
Translated by V. Potapchouck
Rights and permissions
About this article
Cite this article
Khrylchenko, K.Y., Vorontsov, K.V. Optimizing Modality Weights in Topic Models of Transactional Data. Autom Remote Control 83, 1908–1922 (2022). https://doi.org/10.1134/S00051179220120050
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1134/S00051179220120050